No. 1/2015
August 2015
Stakeholder-Oriented Firms:
Evidence from Savings Banks
Charlotte Ostergaard, Ibolya Schindele and Bent Vale
© Charlotte Ostergaard, Ibolya Schindele and Bent Vale 2015. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission, provided that full credit, including © notice, is given to the source. This paper can be downloaded without charge from the CCGR website http://www.bi.edu/ccgr
Social Capital and the Viability of Stakeholder -Oriented Firms: Evidence from Savings Banks
Charlotte Ostergaard1, Ibolya Schindele2, Bent Vale3
1BI Norwegian Business School;2BI Norwegian Business School;3Norges Bank
Abstract. We show that social capital improves the viability of stakeholder-oriented firms operating in competitive markets. Studying exits from the population of Norwegian savings banks after deregulations, we find that banks located in communities with high social capital have a higher probability of survival, but no similar effect exists for commer- cial banks. Norwegian savings banks are collectively governed by their stakeholders and we provide evidence that social capital improves the efficiency of stakeholder governance.
In high social capital areas, banks raise more deposits locally, distribute more of their surplus for altruistic purposes, and operate more locally-focused branch networks.
Keywords: Social Capital, Savings Banks, Stakeholder Governance, Non-profit Firms, Financial Intermediation
JEL Classification: P13, Z13, G21, G30
We thank Steven Ongena (the Co-editor), two anonymous referees, Øyvind Bøhren, Roberta Dessi, Klarita G¨erxhani, Hans Halle, Einar Kleppe, Eirik Kristiansen, Alfred Lehar, Thierry Magnac, Mark Spiegel, Bent Sørensen, Rune Sørensen, Nicolas Van de Sijpe, and seminar participants at the ASSA 2008 Meetings in New Orleans, the CEPR–
CMPO Conference on Institutions and Institutional Change in Bristol, 2008, the EARIE 2008 Conference in Toulouse, the EEA 2008 Meetings in Milan, the EFA 2009 Meetings in Bergen, Central European University, Central Bank of Hungary, Copenhagen Busi- ness School, the Norwegian Savings Banks Association, Toulouse Business School, and the Universities of Amsterdam and Stavanger for comments and suggestions. We benefitted from data supplied by Sigurd Høst, Anne Mari Furuberg at DnB NOR, and Tove Hauge and Turid Ringstad at Redd Barna and support from the Center for Corporate Gover- nance Research at BI. Ostergaard and Schindele are grateful for financial support from the Foundation for the Advancement of Banking Studies. Payam Farshad, P˚al Kollerud, and, especially, Alexandra Coiculescu provided excellent research assistance.
1. Introduction
Governance objectives of stakeholder-oriented firms encompass the welfare of non-shareholder groups such as employees, customers, and the commu- nity at-large. Economists, however, are often sceptical about the practice of stakeholder-oriented governance. Tirole (2001) points out that the provision of adequate incentives for management to maximize the welfare of stake- holders is fraught with difficulties and that conflicting preferences among stakeholders inhibit implementation of the stakeholder ideal. Jensen (2001) argues that firms that attempt to follow the stakeholder ideal will not survive competition from profit-maximizing firms. Yet stakeholder-oriented firms continue to survive in many sectors where they compete with shareholder- owned firms (Hansmann, 1996).
In this paper, we offer a new perspective on the continued existence of stakeholder-oriented firms in competitive industries. We study survival of Norwegian savings banks after deregulation of the banking industry in the mid-1980s subjected savings banks to the full force of competition from commercial banks. Based on Putnam (1993), we hypothesize that in a com- petitive banking market, savings banks’ efficiency and survival depend on the levels of trust and civic engagement (social capital) in the communities where the banks operate.1
The relation between social capital and savings banks’ viability follows from the banks’ organizational form: Savings banks are institutions for collective action by stakeholder groups from the banks’ areas of location because they are governed by representatives of depositors, local govern- ment councils, and employees. They are non-profit firms, prohibited from distributing profits and no group holds residual cash flow rights.2 Savings banks, therefore, have goals other than profit-maximization, and pursue both financial and social objectives. Social capital improves savings banks’ effi- ciency by facilitating collective action among the banks’ stakeholders.3 To the best of our knowledge, our paper is the first to explore the role of social capital in stakeholder-governed firms.
We first show that savings banks survive longer as independent non-profit firms if they are located in communities with high social capital, and that
1 Because trust and civic engagement are key aspects of social capital, in the paper, we will refer to trust and civic engagement as “social capital”. Social capital may be defined as relations between people “that enable participants to act together more effectively to pursue shared objectives for mutual benefit” (Putnam, 1995).
2 See Hansmann (1996) for a definition of non-profit firms.
3 “Efficiency” is here understood in the sense of Putnam (1993), indicating ability to achieve the institutions’ objectives rather than in the more standard sense of profit- maximization, as we explain below.
social capital increases the probability of survival by up to 20 percentage points. This result obtains after controlling for banks’ financial health and a range of local population and market characteristics that may influence banks’ underlying business opportunities, including unemployment, and dis- tributions of age, income, and wealth. The result, therefore, does not follow from a lack of business opportunities and consequent low market penetration by competitors in communities with high social capital. We subsequently show that social capital does not affect the survival of competing commer- cial banks whose organizational form implies that collective action among stakeholders do not play a pivotal role in their governance.
Putnam (1993) argues that social capital improves the efficiency of insti- tutions for collective action by affecting both the demand for and the supply of their services. Citizens in higher social capital communities demand more effective institutional services. An efficient collective institution is in turn responsive to the demands of the community. Consistent with this mecha- nism, we show that savings banks located in high-social capital areas raise a larger fraction of their deposits in their home regions, they operate more locally-focused branch networks, and they distribute more of their profits as gifts to their local communities. Savings banks located in the areas with the highest level of social capital raise 25 percent more deposits locally and donate 27 percent more compared to banks in the poorest social capital areas.
Last, we investigate whether higher social capital savings banks display distinct financial performance. We find that high social capital banks earn lower returns on assets. This lower profitability, however, does not seem to result from poorly performing loan portfolios, but from lower interest rate margins, consistent with our hypothesis that stakeholders’ social goals are prominent in efficiently-governed savings banks’ objective functions. Over- all, our results suggest that social capital plays a role for the efficiency of stakeholder governance and, consequently, the subsistence of nonprofit firms in competitive markets.
Norwegian savings banks compete in the same product markets as for- profit banks and have, since a comprehensive deregulation of branching and credit restrictions in the mid-1980s, faced severe product competition from the branch networks of for-profit banks. The location of most savings banks operating at the time of deregulation is pre-determined in the 19th century.
Consequently, the Norwegian scenario of banking deregulation sets up a quasi-experiment: We observe the disappearance (“exits”) of banks from the population of savings banks from around the time of deregulation (1987) until 2005, and explore which bank and community characteristics determine whether a bank in a given location is able to survive in the competitive
regime. During the sample period, about 50 percent of the banks exit the sample and cease to exist as independent savings banks. Typically, a lack of long-run profitability causes banks to exit by changing their organizational form or accepting acquisition by a larger bank, but exit may also be due to outright failure. At the end of our sample, 102 savings banks compete with 31 other, for-profit banks in Norway.
For every year of the sample, we map out the location of all Norwegian banks’ branches, placing each branch in one of the 433 municipalities and match this data with measures of the level of, among others, social capital in each municipality. We then set up a discrete time survival model and estimate the probability of exit as a function of the level of social capital in the municipalities where the banks operate, controlling for other bank and municipality characteristics. The analysis is conducted with three different measures of trust and civic engagement together with their first principal component: a score of trust from the World Values Survey, households’ news- paper subscriptions, and donations to charity.
Our paper is related to the literature on firms with stakeholder-oriented objectives. Fauver and Fuerst (2006) find that employee representation on German corporate boards increases firms’ market value. Consistent with our approach, Allen, Carletti, and Marquez (2015) argue that stakeholder- oriented firms’ overriding objective is survival in the long term. Ewerhart and Zubrickas (2013) model cooperative banks as optimal intermediaries for depositors with a preference for social equality. Fuertes, Izzeldin, Ongena, and Pappas (2013) show that Islamic banks have a lower risk of failure than conventional banks, suggesting that depositors loyalty instilled by re- ligious beliefs has a role in governance. Illueca and Norden (2013) find that a shift of control from local to regional governments increases risk-taking by state-owned savings banks. Finally, Bøhren and Josefsen (2013) study the financial performance of Norwegian banks and find that savings banks gen- erate returns that are comparable to those produced by commercial banks.
The latter compares the performance of banks with different organizational forms, whereas we focus on non-profit banks and propose a link between social capital and the viability of that organizational form. In addition, our paper is linked with the property rights literature that addresses distinct fea- tures of shareholder versus stakeholder ownership, e.g. Hansmann (1996), Hart and Moore (1998), and Rey and Tirole (2007). Our analysis suggests that social capital may facilitate efficient stakeholder-governance.
Our work is also related to the literature examining the effect of social capital on economic outcomes. Knack and Keefer (1997) and LaPorta et al. (1997) show that countries with more trust have higher economic growth and more efficient judicial systems. Guiso, Sapienza, and Zingales (2004;
2008) document that more trusting individuals are more likely to invest in stocks and make less use of informal credit. Guiso, Sapienza, and Zingales (2009) and Bottazzi, Da Rin, and Hellmann (2015) find that trust enhances cross-border trade and investment.4 Mistrulli and Vacca (2014) show that, in high social capital regions, financial shocks cause smaller increases in the cost of credit. The theme in these papers is how social capital and generalized trust between counter-parties facilitate financial contracting and economic exchange. While we also consider trust concerning peoples’ general beliefs about others’ behavior, we study how those beliefs facilitate collective action.
Finally, our paper is related to a large group of papers on the conse- quences of banking deregulations, including Jayaratne and Strahan (1997), Bertrand, Schoar, and Thesmar (2007), Demyanyk, Ostergaard, and Sørensen (2007), and Rice and Strahan (2010). These papers study how increased competition between banks affects economic and bank-level effi- ciency, such as regional growth, risk sharing, and the terms of finance ex- tended to borrowers. Our paper is similar in that it studies how deregulation forces inefficient banks to exit, but in our setting, banks have social objec- tives, and we study the efficiency of stakeholder governance rather than profit maximization.
The paper proceeds as follows. In Section 2. we discuss the link between community social capital and savings banks’ viability. Section 3. provides a brief overview of the Norwegian banking industry and its development since deregulation. Section 4. describes our data, and Section 5. the methodology.
Section 6. discusses the empirical results, and Section 7. concludes.
2. Trust, civic engagement, and savings banks’ viability
Norwegian savings banks have no owners but are governed by local stake- holders. Depositors and representatives of the municipality governments in the banks’ areas of location each elect an equal number of persons to the Board of Representatives, which in turn elects the Board of Directors and the Control Committee. Depositors from all municipalities where a bank has branches are eligible to be elected and to vote.5 The two boards together
4 In a related vein, Giannetti and Yafeh (2012) provide evidence that cultural distance between professional decision-makers affects lending conditions in the syndicated bank loan market.
5 Statutory law requires that savings banks’ bylaws specify the municipalities from which voters and members of the banks’ governing bodies should be drawn. We randomly sam- pled the 1987-bylaws from 22 savings banks of varying size, to investigate the municipal- ities named in the bylaws. We found that the bylaws named the municipalities in which
elect the CEO.6 The Board of Representatives has a supervisory function vis-a-vis the Board of Directors and will typically set out overall directions for the banks’ strategy. More detailed supervision of the banks’ operations is left to the Control Committee (Bryhni et al. 1985).
Statutory law prohibits savings banks from distributing net profits and requires that residual earnings are used to replenish equity capital or dis- tributed for charitable purposes.7 The non-distribution constraint and the representation of stakeholder groups in banks’ governing bodies imply that the banks are designed to internalize their stakeholders’ interests and have other objectives than the maximization of profits.
Indeed, savings banks have historically pursued both financial and social objectives (Schmidt, 2009). Hansmann (1996) emphasizes that U.S. savings banks were established in response to financial exclusion, providing services to social groups that commercial banks were not prepared to serve at the time. In Norway, social responsibility continues to be an integral part of banks’ strategies, notably in the form of distributions from the gift fund to the banks’ local communities (see Section 3.) .
Essentially, Norwegian savings banks are institutions where representa- tives of different stakeholder groups act collectively to achieve financial and social objectives. Putnam (1993) argues that institutions for collective ac- tion operate more efficiently in communities with high levels of trust and civic engagement in community affairs, with implications for both the de- mand for and the supply of these institutions’ services: “[C]itizens in civic communities [...] demand more effective public service, and they are pre- pared to act collectively to achieve their shared goals.”8 As for the supply of institutional services, he argues that “[a] good [collective institution] not only considers the demands of its citizenry (that is, is responsive), but also acts efficaciously upon these demands (that is, is effective).”9
Prior to deregulation, Norwegian savings banks were protected from com- petition. Deregulation enabled banks to open branches on each others’ turfs, which, in turn, enabled local depositors to move their savings to competing
the bank had branches, and, in a few cases, one (or a couple of) adjacent municipalities without branches.
6 Depositors and local governments together elect 3/4s of the board. The remaining 1/4 is elected by bank employees.
7 A maximum of 25 percent of annual earnings can be set aside in a separate gift fund and distributed for charitable purposes. The rest of the profits is to be retained and reinvested in the bank. In the case of a dissolution, any remaining equity capital must be used to further savings banks business in the bank’s home area. In the case of an acquisition by another savings bank, retained equity is transferred to the merged bank.
8 Putnam (1993), page 182.
9 Putnam (1993), page 63.
banks at low costs if dissatisfied with their local bank. Thus, the financial foundation of banks no longer patronized by their local communities could be undermined, and their equity exhausted over time (Hansmann, 1996).
We hypothesize that in a deregulated environment, social capital mat- ters for the survival of savings banks: In communities with high trust and civic engagement, stakeholder-governed banks operate more efficiently in the sense of Putnam (1993) and, consequently, survive longer. Banks located in such environments will resist mergers, not convert their organizational form, and abstain from the pursuit of business strategies that involve a high probability of default, as these all imply the assumption of control by community-outsiders.10 This is our main hypothesis and we test it in Section 6.. In contrast, social capital should not play a role for the sur- vival of banks of other organizational forms in which collective action by stakeholder groups is less likely to play a role in governance–we test this implication too, using data on Norwegian commercial banks whose owners (shareholders) hold residual cash flow rights (Section 6.4).
The relation between social capital and the viability of Norwegian savings banks may be explored further by examining the underlying mechanisms relating to the demand for and supply of collective institutions’ services.
Building on Putnam (1993), and illustrated in the quotation above, we investigate three mechanisms. First, we propose that in communities with higher levels of social capital, community members patronize local banks by depositing more of their savings in local banks. A stable deposit base is a necessary condition for the long-term survival of banks with limited access to wholesale money markets. We also test whether the local funding of high-social capital banks is more stable in the presence of a common shock to the economy. A large common shock during our sample period is the Norwegian banking crisis which lasted from 1988 to 1993. The crisis is a time period where retail savers faced high uncertainty about the situation of financial institutions and their loyalty would have been tested. Second, the supply side aspect of efficiency suggests that decision-making in high- social capital banks is more responsive to social demands of the community in the bank’s home area. We test whether banks in high social capital areas allocate contribute more to local charity and whether they choose strategies that are more locally oriented by examining the geographical dispersion of their branch network.
10 Conversions of savings banks organizational form have been permitted since 1987 through the issue of a form of equity that introduces owners with residual cash flow rights into the banks’ governing bodies (see Section 3.)
3. A brief history of Norwegian savings banks
At their origin, the savings banks in Norway were a philanthropic project for the bourgeoisie to further the idea of saving among common people. The first banks were established in the larger towns in the 1820s, and thereafter spread to the smaller towns and the countryside. They were organized without owners, had a strong local focus, and part of the surplus was distributed as charity. From the middle of the 19th century the banks incorporated lending as part of their main activities and have, since then, served as an important source of finance for local firms and households (Thue, 2014).
In 1960, 600 savings banks were operating in the country, but economic structural changes prompted a rapid consolidation of the sector, which was more than halved by the mid-1980s.11
Free competition in the Norwegian banking industry was introduced with credit market reforms in the 1980s. Until 1984, quantitative regulations and restrictions on bank branching effectively provided protection for savings banks against entry from outside banks.12 The suspension of restrictions enhanced competition and prompted further consolidation of the industry:
From the time of deregulation till the present, another 50 percent of the independent savings banks agreed to acquisitions or converted its organi- zational form. Savings banks have been able to convert their charter and increase equity capital through the issue of so-called Primary Capital Cer- tificates (PCCs) since 1987. PCCs are residual claims on the banks’ surplus and are traded on the Oslo Stock Exchange.13
Both acquisitions and conversions have been used by savings banks to ac- celerate growth, resulting in regional banks capable of competing with the largest commercial banks. Furthermore, three strategic alliances between independent savings banks were set up during the 1990s.14 Banks within an alliance do typically not operate branches on each others’ home turfs.
They do, however, compete with branches of savings banks from the other
11 In contrast to savings banks in many other countries, Norwegian savings banks are strongly engaged in business lending. At the beginning of our sample, in 1987, loans to businesses made up 31 percent of saving banks’ portfolios, of which 24 percent were com- mercial and industrial loans. In 2005, these fractions were 26 and 23 percent, respectively.
12 To establish new branches, banks were required to obtain approval from the Ministry of Finance, which, through a lengthy process, would consult with the respective local authorities. See also (Norwegian Official Reports , 1992, pp. 66–67).
13 A PCC bank is a hybrid between a commercial bank and a savings bank—it has outside owners with voting and residual cash-flow rights who constitute the largest stakeholder block in the Committee of Representatives.
14 See the Norwegian Savings Bank Association (www.sparebankforeningen.no).
alliances, or savings banks outside the alliances. Hence, savings banks com- pete not just with commercial banks but also with each other.
The banking crisis in 1988-1993 also contributed to the transformation of the industry. From 1988 to 1990, 14 small and some regional banks failed, mostly savings banks. Towards the end of 1990, the situation deterio- rated and the crisis became systemic, forcing the government to establish a governmentally-financed insurance fund.15 None of the failed savings banks were closed. Instead, they were acquired by larger solvent savings banks, or made to sell their devalued equity capital to the Savings Bank Guaran- tee Fund through the issue of PCCs. 15 acquisitions of savings banks and 3 PCC-conversions were the results of these rescue operations, correspond- ing to 20 percent of the savings banks exits in the sample. The pattern of failures during the crisis contains relevant information for our analysis:
Savings banks in low social capital areas are more likely to shift risk and consequently exit early in a deregulated regime.16
Overall, regulatory changes and the consequent transformation of the banking industry resulted in a decrease in the number of savings banks from 191 in 1987 to 102 in 2005. 23 banks converted to PCC-form. The remaining savings banks were acquired by larger banks.
4. Measuring social capital
Trust and civic engagement are key dimensions of social capital (Coleman, 1988; Putnam 1993, 1995, and 2000). We proxy the level of social capital within a community with three different measures that reflect these dimen- sions: (1) a measure of trust from the 1990 World Values Survey, (2) house- hold subscriptions to newspapers, and (3) charity donations. By nature, the measurement of unobservable social capital is not straightforward. For our purposes, proxies for social capital must be available at the municipality or county level, display cross-sectional variation, and not be causally affected by savings banks’ probability of survival. We discuss each measure in turn and refer to the data appendix for the remaining variables used in the re- gressions.
Trust facilitates cooperation towards the implementation of common goals.17 We follow Guiso, Sapienza, and Zingales (2004) and use the WVS
15 At the peak of the crisis in 1991, aggregate loan loss provisions by commercial banks constituted more than 4% of total assets, whereas provisions by savings banks were about 2% of assets.
16 See Moe, Solheim, and Vale (2004) for an account of the Norwegian banking crisis.
17 The social capital literature distinguishes between generalized and personalized trust (see Durlauf and Fafchamps, 2005). The first pertains to people’s preconceptions con-
trust measure to proxy for social capital. Our measure of trust thus indi- cates, on a score of 1–5, the level of trust towards other Norwegians where the score of 5 indicates high trust and the score of 1 high distrust. The variable is available at the county-level.
Interest and knowledge about public issues are necessary conditions for civic engagement in community affairs. Being informed fosters discussion and connectedness among community members. Newspaper readership has been suggested as a measure of civic engagement by Putnam (1993). We use a measure of the average number of newspapers subscribed to by households in each municipality.18
Altruism and volunteering indicate peoples’ willingness to contribute to- wards a general goal at the price of reduced individual consumption and are strongly related to civic engagement.19 Our charity donation measure comes from the annual Norwegian TV charity show—a prime time media event broadcasted nationally on a Sunday in October with the purpose of raising donations for a particular charity organization. On the day of the charity show, door-to-door collections are carried out by volunteers from municipalities all over the country. We construct a municipality-level dona- tion ratio based on the amount raised in day-time door-to-door collections defined as the average donation per unit of income.
Due to its mountainous geography, Norway has a distinct regional char- acter with many small communities and strong regional identities, and the cultural distinctiveness of its regions have been documented by Norwegian sociologists. Since late medieval times, the Southern and Western commu- nities have been more equalitarian communities where forested and agricul- tural land holdings were of homogeneous small size. In contrast, the North- ern, Central, and Eastern regions have had more feudal characteristics with smallholders, forestry and seasonal laborers depending on a few dominant forest owners and large-scale landowners (Rokkan, 1967). Thus, the socioe- conomic heterogeneity of Norwegian communities may be part of the ex-
cerning the behavior of other people not necessarily known to them. The latter concerns the beliefs of one agent about the behavior of another arising from interactions between the two. We focus on generalized trust, so that we are concerned with what might be considered cursory beliefs and generalizations about fellow citizens’ behaviors.
18 Norwegian households’ newspaper consumption per capita is among the highest in the world and the newspaper distribution pattern has a distinct local character (Høst, 2005).
19 Putnam (2000) writes that “ Doing goodforother people, is not part of the definition of social capital. But [...] volunteering and philanthropy and even spontaneously helping are all strongly predicted by civic engagement.”
planation for the geographical pattern in social capital that we document below.20
Figure 1 displays the distribution of the three social capital measures across municipalities. Each map indicates high levels of social capital along the bottom half of the West coast, but otherwise the distributions appear quite dissimilar. This is confirmed by the low cross-correlations between the three measures. Newspaper Subscriptions and Donation Ratio have the highest correlation of 0.31. Trust and Newspaper Subscriptions, respectively Trust and Donation Ratio, have correlations 0.20 and 0.14. By nature, it is not possible to know which proxy comes closest to capturing the true variation in social capital. Therefore, we also run regressions using the first principal component of the three social capital measures.
5. Methodology
We use a discrete-time hazard model to estimate the relations between sav- ings banks’ survival and social capital in the municipalities where the banks operate.21The relevant event is the disappearance of the savings bank as an independent non-profit firm.
To record event occurrence, we divide the time from branching deregula- tion into equal-sized intervals of length one year, with intervalj defined as (j−1, j]. Intervalj = 1 is thus the first year following the date of branch- ing deregulation, 1 January 1984.22 Let T denote the time (years) elapsed from branching deregulation to the observed exit of savings banki, i.e. we have observations onniid random variables, wherenis the number of banks observed at the beginning of interval 1. The hazard rate for bank iin year j is defined as
hij =prob(Ti =j|Ti ≥j, xij), (1) where xij is a (k×1) vector of bank-specific (constant or time-varying) explanatory variables of the event of banki’s exit during observation interval j. that measure the characteristics of bank i and the markets in which it operates, among others, the level of social capital.
20 Accordingly, Putnam (1993) describes cultural heterogeneity and social capital as the product of historical events and proposes that the difference in social capital between the North and South of Italy is rooted in the free city-state experiences of the communities in the North.
21 A detailed derivation of our methodology is provided in an internet appendix accom- panying this paper.
22 We prefer to model the process in discrete rather than continuous time to match the frequency of the exits and the explanatory variables, most of which are available only annually.
We specify a proportional odds logistic model for the hazard rate:
log
"
hij 1−hij
#
= log
"
h0j 1−h0j
#
+β0xij (2)
⇔ hij = 1
1 +e−[θ0j+β0xij] . (3) In (2), the log-odds of the hazard rate for each bank depends linearly on xij and a “baseline” hazard of risk over time, logit(h0j) =θ0j. The base- line hazard is common to all banks and captures the underlying process of consolidation in the banking sector after deregulation.
We specify a functional form forθ0j,
θ0j =α0+α1log(j) +α2[ log(j)]2. (4) In (4), the sign of α1 controls the pattern of duration dependence for the population of savings banks.23 We include a quadratic term to capture the fact that the hazard rate cannot continuously decrease or increase forever since the population of banks is fixed. The form in (4) was chosen based on a preliminary non-parametric estimation of the baseline hazard, with the aim of capturing the “shape” of the process of consolidation in a parsimonious manner, preserving degrees of freedom. As a robustness check, we estimate our main regression using time dummy variables in place of (4).
Brown (1975) and Allison (1982) show how the likelihood function for the event of bank exit can be formulated as the likelihood function for binary dependent variable,yij, equal to one if bankiexits during intervaljand zero otherwise. We thus estimate a logit model withyitas dependent variable and α0, log(j), ( log(j))2, and xij as explanatory variables.
5.1 Construction of duration and explanatory variables
We measure duration of banks’ lifetimes as follows. We collect information on the timing of all acquisitions involving savings banks, on all issues of PCCs, and define the event of exit to take place during the year in which either of these two events occur.24 In the case of acquisitions, target banks are treated as exiting. Essentially all of the mergers that occur during our sample period have clearly defined target and acquiring banks and it is almost always the case that the bank known to be the acquiring bank is
23 Whenα1is negative the hazard rate is monotonically decreasing over time for all banks, and vice versa for positiveα1. Whenα1is zero, the baseline probability of exit is constant for all observation intervals.
24 When exit occurs right at the beginning of a year, i.e. a bank is, say, acquired on 1 January, the event is defined as having taken place during the preceding year.
also the largest.25 New (de novo) savings banks are established during the sample period. We exclude such banks entirely from the analysis as such banks choose location after deregulation has occurred.26
To construct the explanatory variables in (1) we map municipality-level data into bank-specific variables using information on the branch structure of each bank. In each year of the sample, we know the exact location of the banks’ branches. For every bank we can therefore construct a weighted av- erage of the municipality-level variables, where the weights are the fractions of the bank’s branches located in the municipalities.27
For example, let log(POPm) denote the log of the population in municipal-
itymandBRANCHESimdenote the number of branches of bankiin municipal-
itym. We construct the bank-level population variable, “log(Population)i”, as the weighted average of (logged) population size.
log(P opulation)i=X
m
h BRANCHESim
P
mBRANCHESim
·log(POPm)i. (5) The branch structure employed in (5) is the structure that applies at the beginning of each interval (year). Other bank-level explanatory variables, including our measures of social capital, are constructed in a similar manner.
In the estimated model, the explanatory variable of interest is the measure of the level of social capital in the municipalities in which a given bank operates. In addition, we include several other variables in the regression to control for the characteristics of the municipalities, in particular municipality size, the proportion of residents in retirement (proxied by the fraction of the population over 67 years of age), and the education level of the residents in the municipality. Our measures of social capital, are likely to be correlated with these population characteristics—omitting such characteristics might bias our results. Also, donations to charity may be affected by the level of income in a municipality. We therefore scale the charity donation measure by average (gross) personal income in the municipality.
Competition from other banks is also likely to affect the survival of sav- ings banks. We include in our regressions a bank-specific measure of the
25 Except in one case, a new bank, Sparebanken Sogn og Fjordane, was formed by a merger of eight smaller banks. In this case, however, one bank comprised 60 percent of all bank assets in the merger, and we define that bank to be the de-facto acquiring bank.
26 Only four new savings banks have been founded during our sample period (adding only 7 observations to the sample). Including these observations makes no difference to the results.
27 This calculation implicitly assumes that a bank’s branches are all of equal size. The assumption is necessary because data on the distribution of bank assets on municipalities do not exist.
degree of competition a given bank faces from other banks, which we mea- sure in alternative ways. Our preferred measure, “bank asset competition”, captures the average weighted market share of competing banks in munic- ipalities in which a given bank has branches. We proxy market share by total assets assuming that all branches of a given bank are of similar size.28 The alternative competition measures; the number of competing banks, the number of competing banks’ branches, the number of competing large banks (size above the 90th percentile), and the number of competing commercial banks respectively, are computed in a similar manner. Importantly, we al- ways compute the bank market competition measures from information on all municipalities and all banks in the Norwegian banking industry. Our com- petition measures therefore reflect the actual competition a bank is exposed to from allother banks.
We also want to ascertain that surviving high-social capital banks are not located in areas void of business opportunities. Many of our control variables capture aspects of banks’ investment opportunity set (e.g. bank competition, unemployment, population age), but we still include a dummy variable equal to one if a bank is the only bank present in the municipality of its headquarter in the regression.
Furthermore, we include two measures of bank characteristics at the be- ginning of the sample; the equity capital ratio and bank assets in 1987. The suggestion of Hansmann (1996) that savings banks die only slowly because they are not under pressure to generate economic profits, would suggest that a bank can survive in a competitive regime for a longer period of time if it starts out with a considerable level of capital. It is also possible that bank size matters for the probability of survival. Large banks typically have more diversified portfolios, which may improve their risk-return tradeoff, and make them less susceptible to local economic shocks. Bank size and capitalization are, through accounting identities, causally affected by a bank’s continued survival and therefore we use only the 1987-values of these two variables.
Finally, we include control variables for the level of economic activity measured by average personal income and the rate of unemployment, lagged one period. Bank lending may lower local unemployment, and we control for this by including the lagged rate of unemployment. In general we collect municipality level data for as many years of the sample period as possible but statistics are not always available for every year. In such cases, we construct
28 For a given bank, we compute the asset competition it faces as the weighted sum of assets held by competing banks in each municipality similar to (5).
a step-wise variable in accordance with the years of information that are available.29
As a final test of robustness, we run our main regressions taking into account the pattern of failed banks during the banking crisis. In particular, for a failed bank, we determine the year of exit as the first year in which it receives capital from the savings banks guarantee fund.30 This redefinition effectively shifts the distribution of exit dates towards the beginning of the sample and causes more tied observations, which may potentially reduce identification.
6. Results and discussion
6.1 Descriptive statistics
Table 1 summarizes the structure of the Norwegian banking sector in 1987 and 2005 for savings banks, commercial banks and PCC banks respectively.
Common for savings banks and commercial banks is a consolidation in both number of banks and number of branches. The number of non-profit savings banks drops from 191 in 1987 to 102 in 2005 compared to a decrease in the population of commercial banks from 22 to 3 and an increase in the popu- lation of PCC banks from 0 to 23. The average savings bank is considerably smaller than the average commercial or PCC bank. Savings banks have, on average, 7.6 branches in 1987 and 3.4 in 2005, whereas the average commer- cial bank has around 30 branches and a PCC bank has close to 20 branches.
The table also shows that savings banks that have remained independent during the sample period tend to be much smaller than savings banks that have converted their organizational form, consistent with growth being a driver of PCC-conversions. PCC banks have relatively more branches in the populous municipalities. In 1987, 28.8 percent of savings bank branches and 5.8 percent of commercial and PCC bank branches are located in municipal- ities with below-median population. In 2005, the figures are 34.2 for savings banks and on average 19.5 percent for the two other bank types. Hence, it is not the case that savings banks survive because they are predominantly located in municipalities with few inhabitants. Overall, the figures illustrate that competition in the banking market has sharpened considerably since deregulation, also in the smaller municipalities.
29 The data appendix, Appendix A, contains a detailed description of the construction of all variables.
30 The guarantee fund is a private risk-sharing arrangement among the savings banks and a draw on the fund is formally not an exit but a private capital infusion.
Figure 2 contrasts the geographical distribution of savings bank branches in 1987 and 2005 with the corresponding distribution of commercial and PCC banks. The plots suggest that the competition from for-profit banks intensified over the sample period. The dilution of savings banks has oc- curred all over the country but has been especially strong in the northern part.
In Table 2, we display bank-level statistics for the social capital variables which display significant variation cross-sectionally. The low variation of Trust is likely due it being measured at the county level, whereas the other variables are measured at the municipality level. For the remaining variables used in the regressions, we refer the reader to the Appendices where we describe the construction and data sources of control variables and provide basic descriptive statistics.
Table 3 provides a summary of the annual number of exits from our sam- ple of savings banks from 1987 and onwards. The second column shows the number of savings banks present in the beginning of a given year and the third column gives the number of bank exits during each year. Out of the 191 savings banks at the beginning of the sample period, 102 savings banks survive until the end of the sample. The last two columns in the table state estimated Kaplan-Meier survival probabilities and interval hazard rates. The hazard probability is highest in the earliest years of the sample, around 7 percent, and then falls to a lower level of a few percent. It is not mono- tonically decreasing over time: Exits are clustered in the years right after deregulation and at the end of the 1990s.
6.2 Logit regressions of the probability of exit
Table 4 shows the results from logit regressions of the hazard rate on a baseline hazard and explanatory variables. Models (1)–(5) employ the con- tinuous social capital variables described in Section 4., whereas Models (6)–
(10), replace the continuous variables with dummy variables indicating an above-average, “high”, level of social capital.31 The two specifications are alternative ways of modeling the effect of social capital. In the continuous variable-specification, the effect of social capital is assumed to monotonically increasing in the value of the variable. This may be a quite rigid restriction to impose on the data. Given the nature of social capital, one may prefer to treat it as a categorial variable that only distinguishes between high and low levels. We therefore show results with both specifications.
The estimated Models (1)-(4) and (6)-(9) show that all three measures of social capital have a significant and negative effect on the hazard rate, that
31 The average is taken across all municipalities.
is, savings banks’ probability of exit in a given period is lower when banks’
branches are located in municipalities with a high(er) level of social capital.
The effects are all significant at the five percent level or less in both model specifications.
To illustrate the interpretation of the estimated coefficients, consider first the estimated baseline hazard function, α0+ln(j) +ln(j)2. In period one, i.e. 1987,j equals 1 and the baseline hazard reduces toα0. In Model (2), for example, the estimated value ofα0 is positive and equals 4.66, which implies that the odds, (1−hh ), in period one exceeds 1—the baseline probability of exit is higher than the probability of survival. One can compute that the baseline probability of exit in period one equals 0.991, or 99.1 percent.32 The negative sign of the estimated coefficient on Newspaper Subscriptions (-1.33) then implies that a bank with a value of Newspaper Subscriptions equal to 1, has a 94.7 percent probability of exit in period one assuming for simplicity that the values of all other variables are zero.33That is, depending on their signs, the coefficient of the explanatory variables shift the baseline hazard up or down, in the scale of logit-hazard.34 The estimated signs of the coefficients of the second and third term in the baseline hazard function imply that the probability of exiting over time is bell shaped, increasing at first but then falling over time. The estimated joint effect of these two terms is statistically significant at a level less than 1 percent (LR-Test 2).
The estimated effect of banks’ equity ratio at the onset of the deregu- lated regime is also negative and statistically significant at a level below 1 percent—capitalization is clearly an important determinant of the viability of non-profit banks, and this makes economic sense since capital improves a bank’s ability to withstand shocks.
In Columns (5) and (10) we run a horse race between the three measures of social capital. The estimates and levels of significance remain remarkably unchanged, indicating that each measure contains independent information and is picking up different aspects of social capital.
Of the other explanatory variables included in the regression, several are significant at the 5 or 10 percent level. The measure of business risk, Frac- tion of C&I Loans in 1987, is highly significant, reflecting, as predicted, that banks carrying relatively more risk on their books in the years just prior to the crisis, were more likely to exit. More intense competition increases the probability of exit as would be expected—Bank Asset Competition is
32 h= 0.991 solves ln(1−hh ) = 4.66.
33 From ln(1−hh ) = 4.46-1.33.
34 Notice that the probability of exit in 1987 for a given bank depends on its value of all the variables in the model and these will in general not be equal to zero as the above calculations assume.
significant at levels around 10 percent. Large municipality size (Population) lowers the probability of exit, which may reflect the existence of underly- ing business opportunities. Many of the savings banks that have pursued a growth strategy after deregulation are headquartered in the more densely populated regional centers and have been acquiring other banks in mergers.
Only Bank in Home Municipality has a positive sign, implying that being a single bank in an area increases the probability of exit. The variable is insignificant, but its sign reflects that being a single bank in an area does not automatically increase lifetime duration. Hence, our results do not re- flect that savings banks survive simply because they are located in areas without competition. Banks that operate without competition from other banks are probably more likely to be situated in a region void of business opportunities, which in turn, increases the likelihood of exit. Population over 67 years is significant at levels around 15 percent with a positive sign, that is, we do not find evidence that non-profit banks located in communities with a less active population are able to survive longer. In fact, we find evidence of the opposite. Higher Median Income increases the probability of exit, but the coefficient estimate is borderline significant. The estimated coefficients of Total Assets, Education and Unemployment are all insignif- icant, as are State and Municipality Owned Enterprises and Population in Urban Settlement. Finally, we observe that the R-squared is of similar size in all regressions.
In Table 5 we re-estimate the two specifications replacing the parametric log-baseline hazard function with a dummy variable for each period j in which at least one bank exit occurs. The latter specification may capture time-varying macroeconomic developments better than the models with the log-baseline hazard, but, on the other hand, entails a loss of degrees of freedom given that we estimate 19 additional parameters. In fact, in four periods the baseline hazard is inestimable because no bank exit occurs in those years. The results are quite similar to those of Table 4. The estimated coefficients are of the same sign and similar magnitude, and we conclude that our results are robust to alternative specifications of the effect of time.
However, because the parametric baseline specification performs as well as the general dummy-variable specification, but is more parsimonious and can be estimated for all periods, we conduct the rest of our analysis with the log-baseline function.
To get a sense of the economic importance of our results, we use Model (1) in Table 4 to estimate the marginal effect of a discrete change in the value of Trust in the year of 1987, assuming that all other explanatory variables are held at their mean values. When the average level of Trust increases from its
minimum value of 3.92 to its maximum of 4.33, the estimated probability of exit decreases by 6.6 percentage points for the average bank. In the middle of the sample period, 1997, the probability decreases by 2.2 percentage points, reflecting that the probability of exit is estimated to be highest in the begin- ning of the period. For Subscriptions and Donation, Models (2) and (3), the estimated marginal effects are –9.7 and –19.0 percent in 1987, and –3.0 and –4.3 percent in 1997. If one instead considers a discrete increase in Trust of one standard deviation around the mean (from 1/2 standard deviation below to 1/2 above), the corresponding decrease in the probabilities of exit are –1.6 in 1987 and –0.5 percentage points in 1997. For Subscriptions the decrease in probabilities are –1.7 and –0.5, and for Donation –3.3 and –0.7 percentage points respectively. Clearly, for a hypotheticalaverage bank, the economic importance of social capital is considerable when we compare the minimum and maximum values of social capital observed in the sample, but smaller if we look at a smaller change in social capital of one standard-deviation. This suggests that average banks that operate in markets with an average level of social capital experience a relatively modest effect of it. However, banks that operate in communities with above-average social capital experience a markedly improved probability of survival.
The importance of social capital becomes further pronounced when one considers its interaction with other key variables. Economic intuition would suggest that a bank’s capitalization at the time of deregulation should be one of the most important factors for survival, since well-capitalized non- profit firms may continue to survive for long periods of time even if they operate with losses (Hansmann, 1996). Indeed, our estimates corroborate the importance of a bank’s 1987-level of equity. The estimated marginal effect of changes in the ratio of equity capital in Models (1)–(3) is consid- erable. In 1987, a discrete change in Equity Ratio from its minimum to its maximum level, decreases the probability of exit by 43.8, 41.3, and 51.2 per- centage points according to Models (1), respectively (2) and (3), holding all other explanatory variables at their means. Our estimates therefore suggest that social capital may work as a substitute to equity capital: Banks that enjoy the patronage of the local community may operate with lower levels of capital. Generally, smaller banks, including savings banks, tend to op- erate with considerably more equity than larger banks due to their lack of access to wholesale funding (and, possibly, lack of implicit government guar- antees due to their smaller systemic importance) but higher capitalization is costly. Banks in high social capital communities may enjoy a competitive advantage.
In Figure 3, we illustrate this interaction between social capital and equity capital. We depict the estimated effect of social capital on the probability
of exit for different values of Equity Ratio in 1987, using the estimates of Models (1)–(3). All other explanatory variables are held at their mean values.
The plots show that for a (hypothetical) average bank with Equity Ratio equal to the minimum ratio observed in our sample, social capital has a markedly higher effect on the probability of exit compared with a bank with average social capital. On the other hand, social capital has almost no effect on the survival probability of a bank with the maximum observed equity ratio.
6.3 Robustness
Table 7. shows regression results with alternative measures of bank market competition. The regression specification is similar to Models (1)–(4) in Ta- ble 4. As can be seen, the estimated coefficients on Trust, Subscriptions, Donation, and Principal Component are all robust to different measures of competition. The estimated coefficients on the competition measures them- selves are all insignificant at conventional levels and less significant than our preferred measure of competing, Bank Asset Competition, used in Table 4.
It is interesting, however, that the sign of the competition measure in Mod- els (9)–(12), CB Branch Competition which measures competing branches of commercial banks, changes to negative and is close to being significant, in- dicating that stronger competition from commercial bankslowersthe prob- ability of exit. This seems to indicate some market segmentation, consistent with customers of savings banks having a preference for the non-profit or- ganizational form. The insignificance of the results, however, provides only suggestive evidence for such an effect.
As a second test of robustness, we run our main regressions redefining the timing of bank exits during the banking crisis. Thus far in the analysis, all banks, including financially troubled banks, have been set to exit in the year in which they were acquired or issued PCCs. In Table 7 we redefine the year of exit for troubled banks as the first year in which they receive capital from the savings banks guarantee fund. That year typically precedes the year of acquisition or a PCC emission. The savings banks guarantee fund is a private insurance fund and, with the exception of two banks, savings banks did not receive government emergency funding during the crisis.35 Table 7
35 The two exceptions, Sparebanken Midt-Norge and Sparebanken Rogaland, received funding from the guarantee fund in 1991 when the fund was forced to borrow from the government. Hence, in practice, for these two banks, the year of private and public emer- gency funding coincides. Other troubled savings banks drew on the guarantee funding before it ran out of money. Further information on capital infusions from the savings bank guarantee fund may be found in Moe, Solheim, and Vale (2004), ch. 6.
displays the results of the redefined survival regressions. The results show that the effect of social capital is robust to this change in specification.
Our main survival model specification does not contain time-varying bank performance measures, only measures of performance in 1987. This choice is for precautionary reasons: Contemporary performance measures may suffer from rate dependence, i.e. performance may be affected by the risk of exit, which in turn prevents causal inference. For example, the risk of exit may be determined by a bank’s business strategy (growth vs. local focus), and the risk and return associated with a particular business strategy will obviously impact performance.36
On the other hand, omitted variables may lead to biased estimates of social capital’s effect on exit. This would be the case if bank performance is correlated with the risk of exit as well as social capital. For example, bank performance may be correlated with the wealth of the bank’s customer base, which again, may be correlated with social capital. Although such a link is less straightforward, it cannot be ruled out, and we therefore include numerous time-varying demographic characteristics of a bank’s local area as control variables in the regressions.
Although we prefer not to include time-varying bank performance mea- sures in our regressions, we nevertheless present results with such measures included in Table 8 as a third test of robustness. To help alleviate rate de- pendence problems, we lag the measures one period. We include a measure of return (Return-On-Assets, ROA) and risk (Past Due Loans), as well as the time-varying Equity Ratio. We already know from the results in Table 4, that capitalization is the economically most important driver of banks’ longevity.
Because past due loans is only available from 1992, we present results both with and without that variable. In Models (1)–(4), we control for ROA and the equity-ratio.37 Although the coefficient on Trust is no longer significant at the 5 percent level, which may be due to the fact that it varies only at the county level, all estimates of social capital’s effect on exit remain negative and, in the case of the other two measures, significantly different from zero, although the estimated coefficient on Donation appears a bit large. Despite the insignificance of Trust, the estimate corresponding to the first principal component remains strongly significant. When we include Past Due Loans in columns (5)-(8), the number of observations falls considerably, as all ob- servations prior to 1992 are now dropped from the regression. This works to
36 Banks that pursue a growth strategy are more likely to convert their organizational form, whereas banks that maintain a local focus are less likely to convert organizational form or agree to be acquired.
37 ROA is not available in all years for all banks, hence the number of observations is somewhat below that of Table 4.
our detriment, as most exits occur in the early sample period, but the results are qualitatively unchanged compared with columns (1)–(4). We therefore conclude that the estimated effect of social capital is robust to the inclusion of time-varying bank performance measures.
6.4 Further explorations
In this section we present some further results in support of our hypothesis.
The link between social capital and saving banks’ survival works through the banks’ non-profit organizational form. It follows that the longeviety of banks that operate in the same market, but with a different organizational form, should not display a similar dependence on social capital. To test this implication, we conduct a duration analysis also of Norwegian commercial banks.
In 1987, 22 commercial banks exist in Norway. Of these, all but three banks exit by acquisition in either domestic (15 banks) or foreign (4 banks) mergers during the sample period. Table 9 displays the survival table of the 22 commercial banks. With this number of commercial banks, it makes little sense to specify a survival model for commercial banks similar to that of Table 4, because that would require estimation of 19 parameters and the degrees of freedom are simply to few. Instead, we estimate a parsimonious survival model of the form
logit(hij) =
k=4
X
k=1
αkDk,ij+α1log(j) +α2[ log(j)]2, (6) where the first term captures four different levels of social capital (“Low”,
“Middle-Low”, “Middle-High”, and “High”) corresponding to the quartiles of the distribution of social capital over all municipalities, i.e. (6) estimates a different rate of exit for each level of social capital. Dk,ij is a indicator variable equal to one if the social capital of bankiin yearj belongs to level k, and the last two terms capture the time variation in the rate of exit. This specification allows a straightforward estimation of how each level of social capital affects bank survival, as captured by the estimatedαks. We estimate the model for savings banks and commercial banks in turn, measuring social capital by the first principal component.
The results are presented in Table 10. The columns “Numbers of banks in 1987” indicate how many savings banks, respectively, commercial banks, are located in the municipalities in each quartile at the beginning of the sample.
There is a clear difference in the location of the two types of banks. Whereas the distribution of savings banks is skewed towards high social capital areas, commercial banks are mainly located in the low social capital areas. As a
case in point, only one commercial bank is located in the top group. This is prima facie evidence that social capital matters only for the non-profit organizational form. The table displays the point estimates of the αks and the corresponding hazard rates. It is evident that the probability of exit is decreasing in social capital for the savings banks, whereas it appears to be rising for commercial banks (the “High” level coefficient cannot be estimated with only one observation).
As a second test, we pool the observations of savings and commercial banks and estimate a survival model of the form in Table 4, interacting the social capital measures with an indicator variable for the commercial bank form. We wish to test whether the effect of social capital on exit varies with organizational form. A natural specification would be to add an interaction term between social capital and a dummy for organizational form, SCij · DiCB, to the regression. This specification, however, cannot be conducted with continuous social capital measures because these vary very little over time and the interaction term will be almost perfectly correlated withDCBi , creating a problem of multicollinarity.38
We therefore rely on the dichotomous-social capital specification to test an interaction with organizational form. The model is
logit(hij) =α0+β0DiCB+β1SCijhigh+β2DCBi ·SCijhigh+δZij, (7) whereDCB and SChigh are dummy variables for the commercial bank form and high-level social capital banks respectively, and Zit is a vector of the remaining control variables from Table 4 including the log-baseline hazard (α1log(j) +α2[ log(j)]2).39 Table 11 shows the results. The coefficient on the social capital dummy,β1, measures the differential effect of high social capital (as opposed to low) on savings banks’ probability of exit. As before, this effect is negative and strongly significant (except for Donations which is borderline significant). The sumβ1+β2 is the corresponding differential effect of social capital for commercial banks. We test the hypothesis that β1+β2 = 0, that is, that there is no difference in the probability of exit for commercial banks located in high vs. low social capital areas. As displayed in the table, we are unable to reject the null at conventional levels of significance (p-values are 0.39 and higher). We conclude that the effect of social capital does not seem to exist for banks where stakeholders do not play a pivotal role in governance.
38 The correlations between the two terms lie between 0.95 and 0.99 for the different measures of social capital.
39 In this specification, the correlations betweenDiCB·SCijhighandDiCBrange from 0.39 to 0.59.
An alternative test of our hypothesis, is to explore the notion of Put- nam (1993) that social capital matters because it improves the efficiency of collective institutions (see discussion in Section 2.). Compared with low social capital banks, banks in high social capital areas should obtain more deposits from the areas where they have branches, they should distribute more of their profits to the local community as gifts and should employ a more locally-oriented business strategy. In addition, we examine whether high social capital banks have more stable deposit-base in their home region.
We run panel regressions of the above three characteristics on the level of social capital of bank iincluding year dummies and the set of control vari- ables from Table 4. Table 12 shows the results, using the first principal com- ponent measure of social capital. The regressions in the first three columns estimate the effect of social capital on the fraction of deposits raised in the banks’ home regions. This information is available in the bank call reports at the regional level, of which there are 23, and we define a bank’s local area as the region in which its head-quarter is placed. We show results with and without controls for time-varying bank performance measures. The effect of social capital is positive and significant at the 1 percent level, and robust to the inclusion of bank performance variables. The marginal effect of 21.38 implies that a one standard deviation increase in social capital translates into a 4.3 percentage point increase in local deposits, whereas the difference between the minimum and maximum level of social capital translates into an increase of 25.2 percent.40 We measure the distribution of profits to the bank’s local community by the fraction of annual surplus paid out as gifts or put aside in the gift fund for future distribution. The results shows that gifts increase in social capital. The coefficient of 0.23 implies that a one standard deviation increase in social capital increases distributions by 4.6 percentage points. The difference between banks with the minimum and the maximum level of social capital corresponds to a 27.1 percent increase. Finally, we consider the effect of social capital on the number of municipalities in which a bank has branches and the results show that higher social capital banks are present in fewer municipalities, consistent with them having a more local focus. A one standard deviation increase in social capital lowers the presence by 0.3 municipalities, the difference between minimum and maximum social capital lowers presence by 1.7 municipalities. These marginal effects should be seen in relation to the average presence which is 2.5 municipalities.
Lastly, we consider whether local depositors in high social capital-areas display more loyalty towards local savings banks. We first compare the sam- ple variation of local deposit finance (as defined in Table 12) in banks with
40 From 0.2× 21.38 and (3.4-2.22)×21.38 respectively, cf. Table 2.
